Artificial Intelligence Nanodegree

Convolutional Neural Networks

Project: Write an Algorithm for a Dog Identification App

Why We're Here

In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).

Sample Dog Output

In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!

The Road Ahead

We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.

  • Step 0: Import Datasets
  • Step 1: Detect Humans
  • Step 2: Detect Dogs
  • Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
  • Step 4: Use a CNN to Classify Dog Breeds (using Transfer Learning)
  • Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)
  • Step 6: Write your Algorithm
  • Step 7: Test Your Algorithm

Step 0: Import Datasets

Import Dog Dataset

In the code cell below, we import a dataset of dog images. We populate a few variables through the use of the load_files function from the scikit-learn library:

  • train_files, valid_files, test_files - numpy arrays containing file paths to images
  • train_targets, valid_targets, test_targets - numpy arrays containing onehot-encoded classification labels
  • dog_names - list of string-valued dog breed names for translating labels
In [1]:
from sklearn.datasets import load_files       
from keras.utils import np_utils
import numpy as np
from glob import glob

# define function to load train, test, and validation datasets
def load_dataset(path):
    data = load_files(path)
    dog_files = np.array(data['filenames'])
    dog_targets = np_utils.to_categorical(np.array(data['target']), 133)
    return dog_files, dog_targets

# load train, test, and validation datasets
train_files, train_targets = load_dataset('dogImages/train')
valid_files, valid_targets = load_dataset('dogImages/valid')
test_files, test_targets = load_dataset('dogImages/test')

# load list of dog names
dog_names = [item[20:-1] for item in sorted(glob("dogImages/train/*/"))]

# print statistics about the dataset
print('There are %d total dog categories.' % len(dog_names))
print('There are %s total dog images.\n' % len(np.hstack([train_files, valid_files, test_files])))
print('There are %d training dog images.' % len(train_files))
print('There are %d validation dog images.' % len(valid_files))
print('There are %d test dog images.'% len(test_files))

Using TensorFlow backend.

There are 133 total dog categories.
There are 8351 total dog images.

There are 6680 training dog images.
There are 835 validation dog images.
There are 836 test dog images.

Import Human Dataset

In the code cell below, we import a dataset of human images, where the file paths are stored in the numpy array human_files.

In [2]:
import random
random.seed(8675309)

# load filenames in shuffled human dataset
human_files = np.array(glob("lfw/*/*"))
random.shuffle(human_files)

# print statistics about the dataset
print('There are %d total human images.' % len(human_files))

There are 13233 total human images.

Step 1: Detect Humans

We use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images. OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory.

In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.

In [3]:
import cv2                
import matplotlib.pyplot as plt                        
%matplotlib inline                               

# extract pre-trained face detector
haar_face_cascade = cv2.CascadeClassifier('haarcascades/haarcascade_frontalface_alt.xml')
lbp_face_cascade = cv2.CascadeClassifier('lbpcascade/lbpcascade_frontalface_improved.xml')

# load color (BGR) image
img = cv2.imread(human_files[3])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# find faces in image
faces = haar_face_cascade.detectMultiScale(gray)

# print number of faces detected in the image
print('Number of faces detected:', len(faces))

# get bounding box for each detected face
for (x,y,w,h) in faces:
    # add bounding box to color image
    cv2.rectangle(img,(x,y),(x+w,y+h),(255,0,0),2)
    
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()

Number of faces detected: 1

Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in haar_face_cascade and takes the grayscale image as a parameter.

In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.

Write a Human Face Detector

We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.

In [4]:
# returns "True" if face is detected in image stored at img_path
def haar_face_detector(img_path):
    img = cv2.imread(img_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    faces = haar_face_cascade.detectMultiScale(gray)
    return len(faces) > 0

(IMPLEMENTATION) Assess the Human Face Detector

Question 1: Use the code cell below to test the performance of the face_detector function.

  • What percentage of the first 100 images in human_files have a detected human face?
  • What percentage of the first 100 images in dog_files have a detected human face?

Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.

Answer:

In [5]:
from contextlib import contextmanager
import time

@contextmanager
def Time():
    start = time.time()
    def ellapsed(): return time.time() - start
    yield lambda: ellapsed()
    end = time.time()
    def ellapsed(): return end - start
In [6]:
def detector_assessment(detector):
    """
    Test the performance of the given detector algorithm 
    on the images in human_files_short and dog_files_short.
    """

    human_files_short = human_files[:100]
    dog_files_short = train_files[:100]
    
    file_sets = {'human_files_short' : human_files_short, 
                 'dog_files_short'   : dog_files_short}
    
    for key in file_sets.keys():
        human_count = 0
        with Time() as t:
            for file in file_sets[key]:
                if detector(file): human_count += 1            
        print("{}% detection in the first 100 images in {}".format(human_count, key))
        print("Computation time: {} s \n".format(round(t(), 2)))
In [7]:
detector_assessment(haar_face_detector)

97% detection in the first 100 images in human_files_short
Computation time: 2.65 s 

11% detection in the first 100 images in dog_files_short
Computation time: 14.36 s

Question 2: This algorithmic choice necessitates that we communicate to the user that we accept human images only when they provide a clear view of a face (otherwise, we risk having unneccessarily frustrated users!). In your opinion, is this a reasonable expectation to pose on the user? If not, can you think of a way to detect humans in images that does not necessitate an image with a clearly presented face?

Answer:

  • We could use a 'person detection' algorithm that is not based on face detection.

  • Another strategy would be to rely on the dog_detector: if no dog is detected, we can try to detect a human face. (If none is detected, we output a message saying nothing was detected). This method will give good results since the face detector has very little false negatives.

We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on each of the datasets.

In [8]:
## (Optional) TODO: Report the performance of another  
## face detection algorithm on the LFW dataset
### Feel free to use as many code cells as needed.
In [9]:
# returns "True" if face is detected in image stored at img_path
def lbp_face_detector(img_path):
    img = cv2.imread(img_path)
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
    faces = lbp_face_cascade.detectMultiScale(gray)
    return len(faces) > 0
In [10]:
detector_assessment(lbp_face_detector)

86% detection in the first 100 images in human_files_short
Computation time: 0.76 s 

1% detection in the first 100 images in dog_files_short
Computation time: 6.42 s

The LBP algorithm is twice as fast as the Haar algorithm. Its false positive rate is lower. But its true positive rate is also lower.

Step 2: Detect Dogs

In this section, we use a pre-trained ResNet-50 model to detect dogs in images. Our first line of code downloads the ResNet-50 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories. Given an image, this pre-trained ResNet-50 model returns a prediction (derived from the available categories in ImageNet) for the object that is contained in the image.

In [11]:
from keras.applications.resnet50 import ResNet50

# define ResNet50 model
ResNet50_model = ResNet50(weights='imagenet')

Pre-process the Data

When using TensorFlow as backend, Keras CNNs require a 4D array (which we'll also refer to as a 4D tensor) as input, with shape

$$ (\text{nb_samples}, \text{rows}, \text{columns}, \text{channels}), $$

where nb_samples corresponds to the total number of images (or samples), and rows, columns, and channels correspond to the number of rows, columns, and channels for each image, respectively.

The path_to_tensor function below takes a string-valued file path to a color image as input and returns a 4D tensor suitable for supplying to a Keras CNN. The function first loads the image and resizes it to a square image that is $224 \times 224$ pixels. Next, the image is converted to an array, which is then resized to a 4D tensor. In this case, since we are working with color images, each image has three channels. Likewise, since we are processing a single image (or sample), the returned tensor will always have shape

$$ (1, 224, 224, 3). $$

The paths_to_tensor function takes a numpy array of string-valued image paths as input and returns a 4D tensor with shape

$$ (\text{nb_samples}, 224, 224, 3). $$

Here, nb_samples is the number of samples, or number of images, in the supplied array of image paths. It is best to think of nb_samples as the number of 3D tensors (where each 3D tensor corresponds to a different image) in your dataset!

In [12]:
from keras.preprocessing import image                  
from tqdm import tqdm

def path_to_tensor(img_path):
    # loads RGB image as PIL.Image.Image type
    img = image.load_img(img_path, target_size=(224, 224))
    # convert PIL.Image.Image type to 3D tensor with shape (224, 224, 3)
    x = image.img_to_array(img)
    # convert 3D tensor to 4D tensor with shape (1, 224, 224, 3) and return 4D tensor
    return np.expand_dims(x, axis=0)

def paths_to_tensor(img_paths):
    list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]
    return np.vstack(list_of_tensors)

Making Predictions with ResNet-50

Getting the 4D tensor ready for ResNet-50, and for any other pre-trained model in Keras, requires some additional processing. First, the RGB image is converted to BGR by reordering the channels. All pre-trained models have the additional normalization step that the mean pixel (expressed in RGB as $[103.939, 116.779, 123.68]$ and calculated from all pixels in all images in ImageNet) must be subtracted from every pixel in each image. This is implemented in the imported function preprocess_input. If you're curious, you can check the code for preprocess_input here.

Now that we have a way to format our image for supplying to ResNet-50, we are now ready to use the model to extract the predictions. This is accomplished with the predict method, which returns an array whose $i$-th entry is the model's predicted probability that the image belongs to the $i$-th ImageNet category. This is implemented in the ResNet50_predict_labels function below.

By taking the argmax of the predicted probability vector, we obtain an integer corresponding to the model's predicted object class, which we can identify with an object category through the use of this dictionary.

In [13]:
from keras.applications.resnet50 import preprocess_input, decode_predictions

def ResNet50_predict_labels(img_path):
    # returns prediction vector for image located at img_path
    img = preprocess_input(path_to_tensor(img_path))
    return np.argmax(ResNet50_model.predict(img))

Write a Dog Detector

While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained ResNet-50 model, we need only check if the ResNet50_predict_labels function above returns a value between 151 and 268 (inclusive).

We use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).

In [14]:
### returns "True" if a dog is detected in the image stored at img_path
def dog_detector(img_path):
    prediction = ResNet50_predict_labels(img_path)
    return ((prediction <= 268) & (prediction >= 151))

(IMPLEMENTATION) Assess the Dog Detector

Question 3: Use the code cell below to test the performance of your dog_detector function.

  • What percentage of the images in human_files_short have a detected dog?
  • What percentage of the images in dog_files_short have a detected dog?

Answer:

In [15]:
### TODO: Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.

detector_assessment(dog_detector)

1% detection in the first 100 images in human_files_short
Computation time: 4.72 s 

100% detection in the first 100 images in dog_files_short
Computation time: 4.38 s

Step 3: Create a CNN to Classify Dog Breeds (from Scratch)

Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 1%. In Step 5 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.

Be careful with adding too many trainable layers! More parameters means longer training, which means you are more likely to need a GPU to accelerate the training process. Thankfully, Keras provides a handy estimate of the time that each epoch is likely to take; you can extrapolate this estimate to figure out how long it will take for your algorithm to train.

We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have great difficulty in distinguishing between a Brittany and a Welsh Springer Spaniel.

Brittany Welsh Springer Spaniel

It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).

Curly-Coated Retriever American Water Spaniel

Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.

Yellow Labrador Chocolate Labrador Black Labrador

We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.

Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!

Pre-process the Data

We rescale the images by dividing every pixel in every image by 255.

In [21]:
from PIL import ImageFile                            
ImageFile.LOAD_TRUNCATED_IMAGES = True                 

# pre-process the data for Keras
train_tensors = paths_to_tensor(train_files).astype('float32')/255
valid_tensors = paths_to_tensor(valid_files).astype('float32')/255
test_tensors = paths_to_tensor(test_files).astype('float32')/255

100%|██████████| 6680/6680 [00:54<00:00, 122.34it/s]
100%|██████████| 835/835 [00:06<00:00, 154.77it/s]
100%|██████████| 836/836 [00:06<00:00, 135.85it/s]

(IMPLEMENTATION) Model Architecture

Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:

    model.summary()

We have imported some Python modules to get you started, but feel free to import as many modules as you need. If you end up getting stuck, here's a hint that specifies a model that trains relatively fast on CPU and attains >1% test accuracy in 5 epochs:

Sample CNN

Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. If you chose to use the hinted architecture above, describe why you think that CNN architecture should work well for the image classification task.

Answer:

The architecture below (recommended by Keras for classification with little data) consists in a simple stack of 3 convolution layers with a ReLU activation and followed by max-pooling layers. With the exception of ReLU, this is similar to the architectures that Yann LeCun advocated in the 1990s for image classification.

  • We use the ReLU activation functions to solve the vanishing gradient problem.
  • We use MaxPooling layers to reduce the dimensionality after each convolution layer.
  • We finish with two dense layers. The last layer has 133 nodes since there are 133 possible categories.
  • Since we are constructing a multi-class classification we use softmax as the activation function of the last dense layer.

Compile the Model

In [18]:
from keras.models import Sequential
from keras.layers import Conv2D, MaxPooling2D
from keras.layers import Activation, Dropout, Flatten, Dense

model = Sequential()
model.add(Conv2D(filters=32, kernel_size=3, input_shape=(224, 224, 3)))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Conv2D(filters=32, kernel_size=3, padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))

model.add(Conv2D(filters=64, kernel_size=3, padding='same'))
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=3))

# the model so far outputs 3D feature maps (height, width, features)

model.add(Flatten())  # this converts our 3D feature maps to 1D feature vectors
# model.add(Dense(300))
# model.add(Activation('relu'))

model.add(Dense(133))
model.add(Activation('softmax'))

model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_1 (Conv2D)            (None, 222, 222, 32)      896       
_________________________________________________________________
activation_50 (Activation)   (None, 222, 222, 32)      0         
_________________________________________________________________
max_pooling2d_2 (MaxPooling2 (None, 111, 111, 32)      0         
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 111, 111, 32)      9248      
_________________________________________________________________
activation_51 (Activation)   (None, 111, 111, 32)      0         
_________________________________________________________________
max_pooling2d_3 (MaxPooling2 (None, 55, 55, 32)        0         
_________________________________________________________________
conv2d_3 (Conv2D)            (None, 55, 55, 64)        18496     
_________________________________________________________________
activation_52 (Activation)   (None, 55, 55, 64)        0         
_________________________________________________________________
max_pooling2d_4 (MaxPooling2 (None, 18, 18, 64)        0         
_________________________________________________________________
flatten_2 (Flatten)          (None, 20736)             0         
_________________________________________________________________
dense_1 (Dense)              (None, 133)               2758021   
_________________________________________________________________
activation_53 (Activation)   (None, 133)               0         
=================================================================
Total params: 2,786,661.0
Trainable params: 2,786,661.0
Non-trainable params: 0.0
_________________________________________________________________
In [25]:
model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])

(IMPLEMENTATION) Train the Model

Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.

You are welcome to augment the training data, but this is not a requirement.

In [26]:
# Data augmentation strategy as presented in the keras blogpost

from keras.preprocessing.image import ImageDataGenerator

batch_size = 25

# this is the augmentation configuration we will use for training
train_datagen = ImageDataGenerator(
        rescale=1./255,
        shear_range=0.2,
        zoom_range=0.2,
        horizontal_flip=True)

# this is the augmentation configuration we will use for testing:
# only rescaling
test_datagen = ImageDataGenerator(rescale=1./255)

# this is a generator that will read pictures found in
# subfolers of 'data/train', and indefinitely generate
# batches of augmented image data
train_generator = train_datagen.flow_from_directory(
        'dogImages/train',  # this is the target directory
        target_size=(224, 224),  # all images will be resized to 224x224
        batch_size=batch_size,
        class_mode='categorical')  # since we use categorical_crossentropy loss, we need binary labels

# this is a similar generator, for validation data
validation_generator = test_datagen.flow_from_directory(
        'dogImages/valid',
        target_size=(224, 224),
        batch_size=batch_size,
        class_mode='categorical')

Found 6680 images belonging to 133 classes.
Found 835 images belonging to 133 classes.
In [285]:
from keras.callbacks import ModelCheckpoint  

### TODO: specify the number of epochs that you would like to use to train the model.

epochs = 6

### Do NOT modify the code below this line.

checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.from_scratch.hdf5', 
                               verbose=1, save_best_only=True)

# model.fit(train_tensors, train_targets, 
#           validation_data=(valid_tensors, valid_targets),
#           epochs=epochs, batch_size=batch_size, callbacks=[checkpointer], verbose=1)


history0 = model.fit_generator(train_generator,
                               steps_per_epoch=train_tensors.shape[0] // batch_size,
                               epochs=epochs, 
                               verbose=1, 
                               callbacks=[checkpointer],
                               validation_data=(valid_tensors, valid_targets),
                               validation_steps=train_tensors.shape[0] // batch_size)

Epoch 1/6
266/267 [============================>.] - ETA: 0s - loss: 3.3466 - acc: 0.2304Epoch 00000: val_loss improved from inf to 4.26366, saving model to saved_models/weights.best.from_scratch.hdf5
267/267 [==============================] - 105s - loss: 3.3456 - acc: 0.2304 - val_loss: 4.2637 - val_acc: 0.1257
Epoch 2/6
266/267 [============================>.] - ETA: 0s - loss: 3.1789 - acc: 0.2600Epoch 00001: val_loss improved from 4.26366 to 4.22847, saving model to saved_models/weights.best.from_scratch.hdf5
267/267 [==============================] - 105s - loss: 3.1779 - acc: 0.2601 - val_loss: 4.2285 - val_acc: 0.1222
Epoch 3/6
266/267 [============================>.] - ETA: 0s - loss: 3.0122 - acc: 0.2956Epoch 00002: val_loss improved from 4.22847 to 4.16868, saving model to saved_models/weights.best.from_scratch.hdf5
267/267 [==============================] - 105s - loss: 3.0113 - acc: 0.2959 - val_loss: 4.1687 - val_acc: 0.1281
Epoch 4/6
266/267 [============================>.] - ETA: 0s - loss: 2.8802 - acc: 0.3152Epoch 00003: val_loss did not improve
267/267 [==============================] - 105s - loss: 2.8775 - acc: 0.3160 - val_loss: 4.4270 - val_acc: 0.1401
Epoch 5/6
266/267 [============================>.] - ETA: 0s - loss: 2.7811 - acc: 0.3370Epoch 00004: val_loss did not improve
267/267 [==============================] - 105s - loss: 2.7797 - acc: 0.3377 - val_loss: 4.2700 - val_acc: 0.1629
Epoch 6/6
266/267 [============================>.] - ETA: 0s - loss: 2.6636 - acc: 0.3555Epoch 00005: val_loss did not improve
267/267 [==============================] - 104s - loss: 2.6619 - acc: 0.3554 - val_loss: 4.3741 - val_acc: 0.1461
In [27]:
def plot_history(history):
    """
    plots the evolution the loss the accuracy 
    of the model on the training and validation data 
    as a function of the number of epochs  
    """
    
    plt.figure(figsize=(16, 8))
    
    # Plot accuracy
    plt.subplot(1,2,1)
    plt.plot(history.history['acc'])
    plt.plot(history.history['val_acc'])
    plt.title('Model Accuracy')
    plt.ylabel('accuracy')
    plt.xlabel('epoch')
    plt.legend(['training', 'validation'], loc='upper left')
    
    # Plot loss
    plt.subplot(1,2,2)
    plt.plot(history.history['loss'])
    plt.plot(history.history['val_loss'])
    plt.title('Model Loss')
    plt.ylabel('loss')
    plt.xlabel('epoch')
    plt.legend(['training', 'validation'], loc='upper left')
    plt.show()
In [315]:
plot_history(history0)

Load the Model with the Best Validation Loss

In [28]:
model.load_weights('saved_models/weights.best.from_scratch.hdf5')

Test the Model

Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 1%.

In [29]:
# get index of predicted dog breed for each image in test set
dog_breed_predictions = [np.argmax(model.predict(np.expand_dims(tensor, axis=0))) for tensor in test_tensors]

# report test accuracy
test_accuracy = 100*np.sum(np.array(dog_breed_predictions)==np.argmax(test_targets, axis=1))/len(dog_breed_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)

Test accuracy: 12.4402%

Step 4: Use a CNN to Classify Dog Breeds

To reduce training time without sacrificing accuracy, we show you how to train a CNN using transfer learning. In the following step, you will get a chance to use transfer learning to train your own CNN.

Obtain Bottleneck Features

In [30]:
bottleneck_features = np.load('bottleneck_features/DogVGG16Data.npz')
train_VGG16 = bottleneck_features['train']
valid_VGG16 = bottleneck_features['valid']
test_VGG16 = bottleneck_features['test']

Model Architecture

The model uses the the pre-trained VGG-16 model as a fixed feature extractor, where the last convolutional output of VGG-16 is fed as input to our model. We only add a global average pooling layer and a fully connected layer, where the latter contains one node for each dog category and is equipped with a softmax.

In [32]:
from keras.models import Sequential
from keras.layers import GlobalAveragePooling2D, Dense

VGG16_model = Sequential()
VGG16_model.add(GlobalAveragePooling2D(input_shape=train_VGG16.shape[1:]))
VGG16_model.add(Dense(133, activation='softmax'))

VGG16_model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
global_average_pooling2d_1 ( (None, 512)               0         
_________________________________________________________________
dense_2 (Dense)              (None, 133)               68229     
=================================================================
Total params: 68,229.0
Trainable params: 68,229.0
Non-trainable params: 0.0
_________________________________________________________________

Compile the Model

In [33]:
VGG16_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])

Train the Model

In [312]:
checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.VGG16.hdf5', 
                               verbose=1, save_best_only=True)

history1 = VGG16_model.fit(train_VGG16, train_targets, 
                           validation_data=(valid_VGG16, valid_targets),
                           epochs=20, batch_size=20, callbacks=[checkpointer], verbose=1)

Train on 6680 samples, validate on 835 samples
Epoch 1/20
6500/6680 [============================>.] - ETA: 0s - loss: 11.8756 - acc: 0.1232Epoch 00000: val_loss improved from inf to 9.93021, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 5s - loss: 11.8166 - acc: 0.1277 - val_loss: 9.9302 - val_acc: 0.2347
Epoch 2/20
6660/6680 [============================>.] - ETA: 0s - loss: 9.0487 - acc: 0.3182Epoch 00001: val_loss improved from 9.93021 to 8.96892, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 9.0433 - acc: 0.3184 - val_loss: 8.9689 - val_acc: 0.3198
Epoch 3/20
6500/6680 [============================>.] - ETA: 0s - loss: 8.3755 - acc: 0.4026Epoch 00002: val_loss improved from 8.96892 to 8.68468, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 8.3762 - acc: 0.4022 - val_loss: 8.6847 - val_acc: 0.3581
Epoch 4/20
6660/6680 [============================>.] - ETA: 0s - loss: 8.1564 - acc: 0.4336Epoch 00003: val_loss improved from 8.68468 to 8.68048, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 8.1544 - acc: 0.4338 - val_loss: 8.6805 - val_acc: 0.3653
Epoch 5/20
6660/6680 [============================>.] - ETA: 0s - loss: 8.0355 - acc: 0.4587Epoch 00004: val_loss improved from 8.68048 to 8.55203, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 8.0260 - acc: 0.4594 - val_loss: 8.5520 - val_acc: 0.3892
Epoch 6/20
6500/6680 [============================>.] - ETA: 0s - loss: 7.9378 - acc: 0.4726Epoch 00005: val_loss did not improve
6680/6680 [==============================] - 2s - loss: 7.9559 - acc: 0.4717 - val_loss: 8.6363 - val_acc: 0.3880
Epoch 7/20
6500/6680 [============================>.] - ETA: 0s - loss: 7.8727 - acc: 0.4826Epoch 00006: val_loss improved from 8.55203 to 8.40742, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 7.8757 - acc: 0.4822 - val_loss: 8.4074 - val_acc: 0.4048
Epoch 8/20
6500/6680 [============================>.] - ETA: 0s - loss: 7.7328 - acc: 0.4938Epoch 00007: val_loss improved from 8.40742 to 8.35714, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 7.7393 - acc: 0.4934 - val_loss: 8.3571 - val_acc: 0.4060
Epoch 9/20
6500/6680 [============================>.] - ETA: 0s - loss: 7.5172 - acc: 0.5031Epoch 00008: val_loss improved from 8.35714 to 8.16008, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 7.5295 - acc: 0.5024 - val_loss: 8.1601 - val_acc: 0.4012
Epoch 10/20
6500/6680 [============================>.] - ETA: 0s - loss: 7.2787 - acc: 0.5202Epoch 00009: val_loss improved from 8.16008 to 7.97624, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 7.2789 - acc: 0.5201 - val_loss: 7.9762 - val_acc: 0.4180
Epoch 11/20
6500/6680 [============================>.] - ETA: 0s - loss: 7.1568 - acc: 0.5363Epoch 00010: val_loss improved from 7.97624 to 7.90704, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 7.1407 - acc: 0.5371 - val_loss: 7.9070 - val_acc: 0.4323
Epoch 12/20
6520/6680 [============================>.] - ETA: 0s - loss: 7.0663 - acc: 0.5491Epoch 00011: val_loss improved from 7.90704 to 7.84307, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 7.0661 - acc: 0.5490 - val_loss: 7.8431 - val_acc: 0.4311
Epoch 13/20
6500/6680 [============================>.] - ETA: 0s - loss: 7.0578 - acc: 0.5525Epoch 00012: val_loss improved from 7.84307 to 7.76932, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 7.0411 - acc: 0.5534 - val_loss: 7.7693 - val_acc: 0.4383
Epoch 14/20
6500/6680 [============================>.] - ETA: 0s - loss: 6.9511 - acc: 0.5592Epoch 00013: val_loss improved from 7.76932 to 7.73480, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 6.9574 - acc: 0.5587 - val_loss: 7.7348 - val_acc: 0.4527
Epoch 15/20
6500/6680 [============================>.] - ETA: 0s - loss: 6.9145 - acc: 0.5608Epoch 00014: val_loss improved from 7.73480 to 7.71224, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 6.9070 - acc: 0.5614 - val_loss: 7.7122 - val_acc: 0.4587
Epoch 16/20
6500/6680 [============================>.] - ETA: 0s - loss: 6.8964 - acc: 0.5655Epoch 00015: val_loss improved from 7.71224 to 7.62734, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 6.8718 - acc: 0.5669 - val_loss: 7.6273 - val_acc: 0.4551
Epoch 17/20
6500/6680 [============================>.] - ETA: 0s - loss: 6.7968 - acc: 0.5694Epoch 00016: val_loss improved from 7.62734 to 7.55110, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 6.8096 - acc: 0.5686 - val_loss: 7.5511 - val_acc: 0.4623
Epoch 18/20
6500/6680 [============================>.] - ETA: 0s - loss: 6.7030 - acc: 0.5722Epoch 00017: val_loss improved from 7.55110 to 7.45455, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 6.7001 - acc: 0.5726 - val_loss: 7.4546 - val_acc: 0.4671
Epoch 19/20
6500/6680 [============================>.] - ETA: 0s - loss: 6.6191 - acc: 0.5814Epoch 00018: val_loss did not improve
6680/6680 [==============================] - 2s - loss: 6.6243 - acc: 0.5813 - val_loss: 7.5054 - val_acc: 0.4551
Epoch 20/20
6500/6680 [============================>.] - ETA: 0s - loss: 6.5508 - acc: 0.5838Epoch 00019: val_loss improved from 7.45455 to 7.44424, saving model to saved_models/weights.best.VGG16.hdf5
6680/6680 [==============================] - 2s - loss: 6.5775 - acc: 0.5823 - val_loss: 7.4442 - val_acc: 0.4659
In [316]:
plot_history(history1)

Load the Model with the Best Validation Loss

In [34]:
VGG16_model.load_weights('saved_models/weights.best.VGG16.hdf5')

Test the Model

Now, we can use the CNN to test how well it identifies breed within our test dataset of dog images. We print the test accuracy below.

In [35]:
# get index of predicted dog breed for each image in test set
VGG16_predictions = [np.argmax(VGG16_model.predict(np.expand_dims(feature, axis=0))) for feature in test_VGG16]

# report test accuracy
test_accuracy = 100*np.sum(np.array(VGG16_predictions)==np.argmax(test_targets, axis=1))/len(VGG16_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)

Test accuracy: 47.1292%

Predict Dog Breed with the Model

In [36]:
from extract_bottleneck_features import *

def VGG16_predict_breed(img_path):
    # extract bottleneck features
    bottleneck_feature = extract_VGG16(path_to_tensor(img_path))
    # obtain predicted vector
    predicted_vector = VGG16_model.predict(bottleneck_feature)
    # return dog breed that is predicted by the model
    return dog_names[np.argmax(predicted_vector)]

Step 5: Create a CNN to Classify Dog Breeds (using Transfer Learning)

You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.

In Step 4, we used transfer learning to create a CNN using VGG-16 bottleneck features. In this section, you must use the bottleneck features from a different pre-trained model. To make things easier for you, we have pre-computed the features for all of the networks that are currently available in Keras:

The files are encoded as such:

Dog{network}Data.npz

where {network}, in the above filename, can be one of VGG19, Resnet50, InceptionV3, or Xception. Pick one of the above architectures, download the corresponding bottleneck features, and store the downloaded file in the bottleneck_features/ folder in the repository.

(IMPLEMENTATION) Obtain Bottleneck Features

In the code block below, extract the bottleneck features corresponding to the train, test, and validation sets by running the following:

bottleneck_features = np.load('bottleneck_features/Dog{network}Data.npz')
train_{network} = bottleneck_features['train']
valid_{network} = bottleneck_features['valid']
test_{network} = bottleneck_features['test']
In [38]:
### TODO: Obtain bottleneck features from another pre-trained CNN.

bottleneck_features = np.load('bottleneck_features/DogResnet50Data.npz')
train_Resnet50 = bottleneck_features['train']
valid_Resnet50 = bottleneck_features['valid']
test_Resnet50 = bottleneck_features['test']

(IMPLEMENTATION) Model Architecture

Create a CNN to classify dog breed. At the end of your code cell block, summarize the layers of your model by executing the line:

    <your model's name>.summary()

Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.

Answer:

The data set is small (30 to 50 pictures per breed) and similar to the original training data (ResNet was originally trained on the ImageNet which contains photos of each of the 133 breeds). We can therefore use Resnet50's bottleneck_features as input of a fully connected layer. This is equivalent to training a model where:

  • we slice off the end of the neural network
  • add a new fully-connected layer that matches the number of classes in the new dataset
  • randomize the weights of the new fully connected layer
  • freeze all the weights from the pre-trained network (to avoid overfitting on the small dataset)
  • train the network to update the weights of the new fully connected layer

Since the datasets are similar, images from each data set will have similar higher-level features. Therefore most or all of the pre-trained Resnet50 layers already contain relevant information about the new data set and should be kept.

We use a gap layer to reduce the dimensionality (and therefore the number of parameters and the computation time) of the model.

In [40]:
### TODO: Define your architecture.

from keras.models import Sequential
from keras.layers import GlobalAveragePooling2D
from keras.layers import Activation, Dense

Resnet50_model = Sequential()
Resnet50_model.add(GlobalAveragePooling2D(input_shape=train_Resnet50.shape[1:]))
# Resnet50_model.add(Flatten(input_shape=train_Resnet50.shape[1:]))
Resnet50_model.add(Dense(133, kernel_initializer='random_uniform', bias_initializer='zeros'))
Resnet50_model.add(Activation('softmax'))

Resnet50_model.summary()

_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
global_average_pooling2d_2 ( (None, 2048)              0         
_________________________________________________________________
dense_3 (Dense)              (None, 133)               272517    
_________________________________________________________________
activation_54 (Activation)   (None, 133)               0         
=================================================================
Total params: 272,517.0
Trainable params: 272,517.0
Non-trainable params: 0.0
_________________________________________________________________

(IMPLEMENTATION) Compile the Model

In [41]:
### TODO: Compile the model.

Resnet50_model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])

(IMPLEMENTATION) Train the Model

Train your model in the code cell below. Use model checkpointing to save the model that attains the best validation loss.

You are welcome to augment the training data, but this is not a requirement.

In [319]:
### TODO: Train the model.

checkpointer = ModelCheckpoint(filepath='saved_models/weights.best.Resnet50.hdf5', 
                               verbose=1, save_best_only=True)

history2 = Resnet50_model.fit(train_Resnet50, train_targets, 
                              validation_data=(valid_Resnet50, valid_targets),
                              epochs=20, batch_size=20, callbacks=[checkpointer], verbose=1)

Train on 6680 samples, validate on 835 samples
Epoch 1/20
6620/6680 [============================>.] - ETA: 0s - loss: 1.6153 - acc: 0.6024Epoch 00000: val_loss improved from inf to 0.81474, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 4s - loss: 1.6095 - acc: 0.6031 - val_loss: 0.8147 - val_acc: 0.7533
Epoch 2/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.4468 - acc: 0.8612Epoch 00001: val_loss improved from 0.81474 to 0.71390, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s - loss: 0.4461 - acc: 0.8612 - val_loss: 0.7139 - val_acc: 0.7737
Epoch 3/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.2675 - acc: 0.9130Epoch 00002: val_loss improved from 0.71390 to 0.68442, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s - loss: 0.2673 - acc: 0.9129 - val_loss: 0.6844 - val_acc: 0.8072
Epoch 4/20
6600/6680 [============================>.] - ETA: 0s - loss: 0.1767 - acc: 0.9461Epoch 00003: val_loss improved from 0.68442 to 0.66629, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s - loss: 0.1759 - acc: 0.9463 - val_loss: 0.6663 - val_acc: 0.8012
Epoch 5/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.1276 - acc: 0.9594Epoch 00004: val_loss improved from 0.66629 to 0.63753, saving model to saved_models/weights.best.Resnet50.hdf5
6680/6680 [==============================] - 1s - loss: 0.1279 - acc: 0.9591 - val_loss: 0.6375 - val_acc: 0.8168
Epoch 6/20
6600/6680 [============================>.] - ETA: 0s - loss: 0.0877 - acc: 0.9711Epoch 00005: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0878 - acc: 0.9711 - val_loss: 0.7378 - val_acc: 0.8144
Epoch 7/20
6600/6680 [============================>.] - ETA: 0s - loss: 0.0625 - acc: 0.9805Epoch 00006: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0628 - acc: 0.9804 - val_loss: 0.7327 - val_acc: 0.8072
Epoch 8/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.0472 - acc: 0.9855Epoch 00007: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0471 - acc: 0.9855 - val_loss: 0.7442 - val_acc: 0.8096
Epoch 9/20
6580/6680 [============================>.] - ETA: 0s - loss: 0.0345 - acc: 0.9892Epoch 00008: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0344 - acc: 0.9894 - val_loss: 0.7132 - val_acc: 0.8204
Epoch 10/20
6600/6680 [============================>.] - ETA: 0s - loss: 0.0280 - acc: 0.9920Epoch 00009: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0282 - acc: 0.9919 - val_loss: 0.7642 - val_acc: 0.8168
Epoch 11/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.0199 - acc: 0.9941Epoch 00010: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0199 - acc: 0.9942 - val_loss: 0.7715 - val_acc: 0.8204
Epoch 12/20
6580/6680 [============================>.] - ETA: 0s - loss: 0.0171 - acc: 0.9960Epoch 00011: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0174 - acc: 0.9960 - val_loss: 0.8235 - val_acc: 0.8228
Epoch 13/20
6600/6680 [============================>.] - ETA: 0s - loss: 0.0127 - acc: 0.9973Epoch 00012: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0126 - acc: 0.9973 - val_loss: 0.8073 - val_acc: 0.8228
Epoch 14/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.0116 - acc: 0.9974Epoch 00013: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0116 - acc: 0.9975 - val_loss: 0.8128 - val_acc: 0.8204
Epoch 15/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.0100 - acc: 0.9971Epoch 00014: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0099 - acc: 0.9972 - val_loss: 0.8399 - val_acc: 0.8275
Epoch 16/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.0092 - acc: 0.9980Epoch 00015: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0093 - acc: 0.9979 - val_loss: 0.8660 - val_acc: 0.8144
Epoch 17/20
6600/6680 [============================>.] - ETA: 0s - loss: 0.0085 - acc: 0.9977Epoch 00016: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0084 - acc: 0.9978 - val_loss: 0.8812 - val_acc: 0.8251
Epoch 18/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.0064 - acc: 0.9982Epoch 00017: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0076 - acc: 0.9981 - val_loss: 0.8736 - val_acc: 0.8251
Epoch 19/20
6600/6680 [============================>.] - ETA: 0s - loss: 0.0071 - acc: 0.9982Epoch 00018: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0071 - acc: 0.9982 - val_loss: 0.9239 - val_acc: 0.8251
Epoch 20/20
6620/6680 [============================>.] - ETA: 0s - loss: 0.0051 - acc: 0.9983Epoch 00019: val_loss did not improve
6680/6680 [==============================] - 1s - loss: 0.0051 - acc: 0.9984 - val_loss: 0.9432 - val_acc: 0.8228
In [320]:
plot_history(history2)

(IMPLEMENTATION) Load the Model with the Best Validation Loss

In [42]:
### TODO: Load the model weights with the best validation loss.

Resnet50_model.load_weights('saved_models/weights.best.Resnet50.hdf5')

(IMPLEMENTATION) Test the Model

Try out your model on the test dataset of dog images. Ensure that your test accuracy is greater than 60%.

In [43]:
### TODO: Calculate classification accuracy on the test dataset.

# get index of predicted dog breed for each image in test set
Resnet50_predictions = [np.argmax(Resnet50_model.predict(np.expand_dims(feature, axis=0))) for feature in test_Resnet50]

# report test accuracy
test_accuracy = 100*np.sum(np.array(Resnet50_predictions)==np.argmax(test_targets, axis=1))/len(Resnet50_predictions)
print('Test accuracy: %.4f%%' % test_accuracy)

Test accuracy: 81.8182%

(IMPLEMENTATION) Predict Dog Breed with the Model

Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan_hound, etc) that is predicted by your model.

Similar to the analogous function in Step 5, your function should have three steps:

  1. Extract the bottleneck features corresponding to the chosen CNN model.
  2. Supply the bottleneck features as input to the model to return the predicted vector. Note that the argmax of this prediction vector gives the index of the predicted dog breed.
  3. Use the dog_names array defined in Step 0 of this notebook to return the corresponding breed.

The functions to extract the bottleneck features can be found in extract_bottleneck_features.py, and they have been imported in an earlier code cell. To obtain the bottleneck features corresponding to your chosen CNN architecture, you need to use the function

extract_{network}

where {network}, in the above filename, should be one of VGG19, Resnet50, InceptionV3, or Xception.

In [44]:
### TODO: Write a function that takes a path to an image as input
### and returns the dog breed that is predicted by the model.

from extract_bottleneck_features import *

def Resnet50_predict_breed(img_path):
    # extract bottleneck features
    bottleneck_feature = extract_Resnet50(path_to_tensor(img_path))
    # obtain predicted vector
    predicted_vector = Resnet50_model.predict(bottleneck_feature)
    # return dog breed that is predicted by the model
    return dog_names[np.argmax(predicted_vector)]

Step 6: Write your Algorithm

Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,

  • if a dog is detected in the image, return the predicted breed.
  • if a human is detected in the image, return the resembling dog breed.
  • if neither is detected in the image, provide output that indicates an error.

You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and dog_detector functions developed above. You are required to use your CNN from Step 5 to predict dog breed.

Some sample output for our algorithm is provided below, but feel free to design your own user experience!

Sample Human Output

(IMPLEMENTATION) Write your Algorithm

In [56]:
### TODO: Write your algorithm.
### Feel free to use as many code cells as needed.

import re

BOLD = '\033[1m'
CYAN = '\u001b[36;1m'
BLUE = '\u001b[34;1m'
END = '\033[0m'

def display_prediction(model, img_path, is_dog=True):
    """
    Displays:
    - the ground truth (if the image is that of a dog)
    - the prediction
    - the input image
    
    And, if the prediction is different from the ground truth
    or if the image is that of a human:
    
    - an actual example from the breed predicted
    """

    # -------------- the ground truth --------------

    match = re.search(r"^dogImages\/.*\/[0-9]*\.(.*)\/.*", img_path)
    if match:
        ground_truth = match.group(1)
        print(BOLD + CYAN + "Ground truth \t \t: " + END + ground_truth)
    
    # --------------- the prediction ---------------
    
    predicted_breed = model(img_path)
    print(BOLD + BLUE + "Predicted breed \t: " + END + predicted_breed)

    # --------------- the input image --------------

    plt.figure(figsize=(16, 8))
    
    
    img = cv2.imread(img_path)
    cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    plt.subplot(1,2,1)
    plt.imshow(cv_rgb)
    plt.axis('off')

    # --- actual example from the breed predicted ---
    
    img_path_ex = glob("dogImages/train/*" + predicted_breed + "/*")[0]
    
    
    img_ex = cv2.imread(img_path_ex)
    cv_rgb_ex = cv2.cvtColor(img_ex, cv2.COLOR_BGR2RGB)
    plt.subplot(1,2,2)
    plt.imshow(cv_rgb_ex)
    plt.axis('off')

    plt.title('A real ' + predicted_breed)

    plt.show()
In [57]:
def dog_app(img_path):
    """
    If a dog is detected in the image returns an estimate of the dog's breed. 
    If a human is detected returns an estimate of the dog breed that is most resembling.
    """
    
    if dog_detector(img_path):
        print("That's a cute dog!")
        is_dog=False
    elif haar_face_detector(img_path):
        print("That's a cool human!")
        is_dog=True
    else:
        print("Well... I don't know what that is.")
        return
    
    display_prediction(Resnet50_predict_breed, img_path, is_dog)

Step 7: Test Your Algorithm

In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog's breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?

(IMPLEMENTATION) Test Your Algorithm on Sample Images!

Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.

Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.

In [58]:
def test_app(nb_file=6):
    """
    Tests the app on a few exemples
    """
    
    dog_files_short = list(train_files[:nb_file//2])
    human_files_short = list(human_files[:nb_file//2])
    
    test_files = dog_files_short + human_files_short
    
    for img_path in test_files:
        dog_app(img_path)
In [60]:
test_app(10)

That's a cute dog!
Ground truth 	 	: Kuvasz
Predicted breed 	: Kuvasz

That's a cute dog!
Ground truth 	 	: Dalmatian
Predicted breed 	: Dalmatian

That's a cute dog!
Ground truth 	 	: Irish_water_spaniel
Predicted breed 	: Irish_water_spaniel

That's a cute dog!
Ground truth 	 	: American_staffordshire_terrier
Predicted breed 	: American_staffordshire_terrier

That's a cute dog!
Ground truth 	 	: American_staffordshire_terrier
Predicted breed 	: American_staffordshire_terrier

That's a cool human!
Predicted breed 	: American_water_spaniel

That's a cool human!
Predicted breed 	: English_toy_spaniel

That's a cool human!
Predicted breed 	: Silky_terrier

That's a cool human!
Predicted breed 	: Chesapeake_bay_retriever

That's a cool human!
Predicted breed 	: American_foxhound

Let's test the algorithm on some tricky examples that were mentionned previously:

In [77]:
tricky_tests = ["Brittany", 
                    "Welsh_springer_spaniel", 
                    "Curly-coated_retriever"]

tricky_tests_paths = [glob("dogImages/*/*/" + test + "*")[0] for test in interesting_tests]
tricky_tests_paths += glob("dogImages/*/*/Labrador_retriever*")[:3]
In [78]:
for img_path in tricky_tests_paths:
    dog_app(img_path)

That's a cute dog!
Ground truth 	 	: Brittany
Predicted breed 	: Brittany

That's a cute dog!
Ground truth 	 	: Welsh_springer_spaniel
Predicted breed 	: Welsh_springer_spaniel

That's a cute dog!
Ground truth 	 	: Curly-coated_retriever
Predicted breed 	: Curly-coated_retriever

That's a cute dog!
Ground truth 	 	: Labrador_retriever
Predicted breed 	: Labrador_retriever

That's a cute dog!
Ground truth 	 	: Labrador_retriever
Predicted breed 	: Labrador_retriever

That's a cute dog!
Ground truth 	 	: Labrador_retriever
Predicted breed 	: Golden_retriever

Answer:

I am quite impressed with the prediction! Although they take more time to compute than I expected.

Three improvements:

  • The dataset is very small. A first improvement could be to generate new data with augmentation algorithms.
  • A second improvement could be to train many models using the pre-trained nets (VGG19, Resnet50, InceptionV3, Xception) and to base the predictions on a vote by all the models.
  • A final improvement could be to retrain the last convolution layers to modify the higher level features and adapt them to the dataset.